The optimal way to build speech understanding modules depends on the amount of training data available. When only a small amount of training data is available, effective allocation of the data is crucial to preventing overfitting of statistical methods. We have developed a method for allocating a limited amount of training data in accordance with the amount available. Our method exploits rule-based methods for when the amount of data is small, which are included in our speech understanding framework based on multiple model combinations, i.e., multiple automatic speech recognition (ASR) modules and multiple language understanding (LU) modules, and then allocates training data preferentially to the modules that dominate the overall performance of speech understanding. Experimental evaluation showed that our allocation method consistently outperforms baseline methods that use a single ASR module and a single LU module while the amount of training data increases.

We investigate the impact of traffic on the performance of large-scale NAT (LSN), since it has been attracting attention as a means of better utilizing the limited number of global IPv4 addresses. We focus on the number of active flows because they drive up the LSN memory requirements in two ways; more flows must be held in LSN memory, and more global IPv4 addresses must be prepared. Through traffic measurement data analysis, we found that more than 1% of hosts generated more than 100 TCP flows or 486 UDP flows at the same time, and on average, there were 1.43-3.99 active TCP flows per host, when the inactive timer used to clear the flow state from a flow table was set to 15 s. When the timer is changed from 15 s to 10 min, the number of active flows increases more than tenfold. We also investigate how to reduce the above impact on LSN in terms of saving memory space and accommodating more users for each global IPv4 address. We show that to save memory space, regulating network anomalies can reduce the number of active TCP flows on an LSN by a maximum of 48.3% and by 29.6% on average. We also discuss the applicability of a batch flow-arrival model for estimating the variation in the number of active flows, when taking into account that the variation is needed to prepare an appropriate memory space. One way to allow each global IPv4 address to accommodate more users is to better utilize destination IP address information when mapping a source IP address from a private address to a global IPv4 address. This can effectively reduce the required number of global IPv4 addresses by 85.9% for TCP traffic and 91.9% for UDP traffic on average.

A low complexity log-likelihood ratio (LLR) calculation for high-order amplitude phase shift keying (APSK) signals is proposed. Using proper constellation partitioning together with a look-up table, the number of terms for the comparison of Euclidean distances can be significantly reduced. Compared with the log-sum LLR approximation, the proposed method reduces the computational complexity by more than 65% and 75% for 16-APSK and 32-APSK signals, respectively, with very small bit error rate performance degradation.

In this paper, a new skin detection method using pixel color and image regional information, intended for objectionable image filtering is proposed. The method consists of three stages: skin detection, feature extraction and image classification. Skin detection is implemented in two steps. First, a Sinc function, fitted to skin color distribution in the Cb-Cr chrominance plane is used for detecting pixels with skin color properties. Next, to benefit regional information, based on the theory of color image reproduction, it's shown that the scattering of skin pixels in the RGB color space can be approximated by an exponential function. This function is incorporated to extract the final accurate skin map of the image. As objectionable image features, new shape and direction features, along with area feature are extracted. Finally, a Multi-Layer Perceptron trained with the best set of input features is used for filtering images. Experimental results on a dataset of 1600 images illustrate that the regional method improves the pixel-based skin detection rate by 10%. The final classification result with 94.12% accuracy showed better results when compared to other methods.

We consider wireless secure communications between a source and a destination aided by a multi-antenna relay, in the presence of an eavesdropper. In particular, two cooperation schemes of the relay are explored: cooperative relaying (CR) and cooperative jamming (CJ). We first investigate the transmit weight optimization of CR and CJ, for both cases with and without the eavesdropper's channel state information (ECSI). Then, for the case with ECSI, we derive the conditions under which CR achieves a higher secrecy rate than CJ; for the case without ECSI, we compare the secrecy rates of CR and CJ in high transmit power regimes. Building on this, we propose a novel hybrid scheme in which the relay utilizes both CR and CJ, and study the power allocation of the relay between CR and CJ for maximizing the secrecy rate under individual power constraints. Further, we study the case with imperfect channel state information (CSI) for both CR and CJ. At last, extensive numerical results are provided.

The non-homogeneous Poisson process (NHPP) has been applied successfully to model nonstationary counting phenomena for a large class of problems. In software reliability engineering, the NHPP-based software reliability models (SRMs) are of a very important class. Since NHPP is characterized by its rate (intensity) function, which is known as the software failure rate of NHPP-based SRM, it is of great interest to estimate accurately the rate function from observed software failure data. In the existing work the same authors introduced a Haar-wavelet-based technique for this problem and found that the Haar wavelet transform provided a very powerful performance in estimating software failure rate. In this paper, we consider the application potentiality of a Daubechies wavelet estimator in the estimation of software failure rate, given the software failure time data. We give practical solutions by overcoming technical difficulties in applying the Daubechies wavelet estimator to the real software failure time data.

A novel long-term sub-band entropy (LT-SubEntropy) measure, which uses improved long-term spectral analysis and sub-band entropy, is proposed for voice activity detection (VAD). Based on the measure, we can accurately exploit the inherent nature of the formant structure on speech spectrogram (the well-known as voiceprint). Results show that the proposed VAD is superior to existing standard VAD methods at low SNR levels, especially at variable-level noise.

In this paper, we discuss the stochastic modeling for operational software reliability measurement, assuming that the testing environment is originally different from the user operation one. In particular, we introduce the concept of systemability which is defined as the reliability characteristic subject to the uncertainty of the field operational environment into the model. First we introduce the environmental factor to consistently bridge the gap between the software failure-occurrence characteristics during the testing and the operation phases. Then we consider the randomness of the environmental factor, i.e., the environmental factor is treated as a random-distributed variable. We use the Markovian imperfect debugging model to describe the software reliability growth phenomena in the testing and the operation phases. We derive the analytical solutions of the several operational software reliability assessment measures which are given as the functions of time and the number of debuggings. Finally, we show several numerical illustrations to investigate the impacts of the consideration of systemability on the field software reliability evaluation.

In this letter, we propose a packet loss probability (PLP) assessment method that uses active measurements. Considering the statistical nature of measurement data in a network, we adopt the confidence interval to assess whether the performance of a network complies with a target PLP or not. Using both analysis and simulations, we show that the proposed method can guarantee that the probabilities of erroneous assessments are not more than a given significance level. In addition, we provide a systematic method to determine the number of probing packets needed for statistical assurance by presenting a clear relation between the assessment accuracy and the measurement overhead.

This letter presents a robust receiver using the generalized sidelobe canceller aided with the high-order derivative constraint technique for multicarrier code-division multiple-access (MC-CDMA) uplink against carrier frequency offset (CFO). Numerical results demonstrate the efficacy of the proposed receiver.

In this paper, the correspondence between the weighted line graph and the Mason signal flow graph (MSFG) has been established, which gives an interpretation of a convolutional network code (CNC) over a cyclic network from a different perspective. Furthermore, by virtue of Mason theorem, we present two new equivalent conditions to evaluate whether the global encoding kernels (GEKs) can be uniquely determined by the given complete set of local encoding kernels (LEKs) in a CNC over a cyclic network. These two new equivalent conditions turn out to be more intuitive. Moreover, we give an alternative simple proof of an existing result.

Non-negative matrix factorization (NMF) is widely used for music transcription because of its efficiency. However, the conventional NMF-based music transcription algorithm often causes harmonic confusion errors or time split-up errors, because the NMF decomposes the time-frequency data according to the activated frequency in its time. To solve these problems, we proposed an NMF with temporal continuity and harmonicity constraints. The temporal continuity constraint prevented the time split-up of the continuous time components, and the harmonicity constraint helped to bind the fundamental with harmonic frequencies by reducing the additional octave errors. The transcription performance of the proposed algorithm was compared with that of the conventional algorithms, which showed that the proposed method helped to reduce additional false errors and increased the overall transcription performance.

3 dB bandwidth enhancement of single mode semiconductor lasers is confirmed numerically and experimentally when they are operated by intensity modulated signal light injection. 3 dB bandwidth is enlarged to 2.5 times of resonant frequency. The numerical analysis of rate equations predicts that the bandwidth enhancement is accomplished by the modal gain control of semiconductor lasers with injected intensity modulated signal light through non-linear gain coefficient term.

Performance of CSMA/CA wireless communication is severely affected by hidden terminal (HT) problem that results in failure of carrier sense and causes packet error due to collision. However, no mathematical analysis method for the HT problem has been available that takes into account actual radio environments including both fading and capture effect. This paper presents an analysis method that enables to well predict the probability of successful communication (PSC) and communication efficiency for CSMA/CA unicast communication including the interaction of data and ACK packets. Analysis of the PSC with two-dimensional HT distribution makes it easy to understand the influence of HT location and carrier sense level. Also it is shown that there is considerable difference on the PSC between fading and fading-free environments. The obtained results as well as the proposed analysis method are quite useful in CSMA/CA network design for WLAN and sensor network applications.

This paper presents a response time acceleration technique in a high-gain capacitive-feedback frontend amplifier (FA) for high output impedance sensors. Using an auxiliary amplifier as a unity-gain buffer, a sample-and-hold capacitor which is used for band-limiting and sampling the FA output is driven at the beginning of the transient response to make the response faster and then it is re-charged directly by the FA output. A condition and parameters for the response time acceleration using this technique while maintaining the noise level unaffected are discussed. Theoretical analysis and simulation results show that the response time can be less than half of the case without the acceleration technique for the specified settling error of less than 0.5%.

This paper evaluates the delay of the issue queue in a superscalar processor to aid microarchitectural design, where quick quantification of the complexity of the issue queue is needed to consider the tradeoff between clock cycle time and instructions per cycle. Our study covers two aspects. First, we introduce banking tag RAM, which comprises the issue queue, to reduce the delay. Unlike normal RAM, this is not straightforward, because of the uniqueness of the issue queue organization. Second, we explore and identify the correct critical path in the issue queue. In a previous study, the critical path of each component in the issue queue was summed to obtain the issue queue delay, but this does not give the correct delay of the issue queue, because the critical paths of the components are not connected logically. In the evaluation assuming 32-nm LSI technology, we obtained the delays of issue queues with eight to 128 entries. The process of banking tag RAM and identifying the correct critical path reduces the delay by up to 20% and 23% for 4- and 8-issue widths, respectively, compared with not banking tag RAM and simply summing the critical path delay of each component.

Network-on-chip (NoC) architectures have emerged as a promising solution to the lack of scalability in multi-processor systems-on-chips (MPSoCs). With the explosive growth in the usage of multimedia applications, it is expected that NoC serves as a multimedia server supporting multi-class services. In this paper, we propose a configuration algorithm for a hybrid bus-NoC architecture together with simulation results. Our target architecture is a hybrid bus-NoC architecture, called busmesh NoC, which is a generalized version of a hybrid NoC with local buses. In our BMNoC configuration algorithm, cores which have a heavy communication volume between them are mapped in a cluster node (CN) and connected by a local bus. CNs can have communication with each other via edge switches (ESes) and mesh routers (MRs). With this hierarchical communication network, our proposed algorithm can improve the latency as compared with conventional methods. Several realistic applications applied to our algorithm illustrate the better performance than earlier studies and feasibility of our proposed algorithm.

In this letter a general admission control strategy is proposed and mathematically analyzed. Fractional buffering finely adjusts different QoS metrics allowing them to simultaneously achieve their maximum acceptable values, maximizing system capacity. Fractional buffering also allows the adequate and fair performance comparison among different resource management strategies and/or evaluation scenarios.

When computing resources are consolidated in a few huge data centers, a massive amount of data is transferred to each data center over a wide area network (WAN). This results in increased power consumption in the WAN. A distributed computing network (DCN), such as a content delivery network, can reduce the traffic from/to the data center, thereby decreasing the power consumed in the WAN. In this paper, we focus on the energy-saving aspect of the DCN and evaluate its effectiveness, especially considering traffic locality, i.e., the amount of traffic related to the geographical vicinity. We first formulate the problem of optimizing the DCN power consumption and describe the DCN in detail. Then, numerical evaluations show that, when there is strong traffic locality and the router has ideal energy proportionality, the system's power consumption is reduced to about 50% of the power consumed in the case where a DCN is not used; moreover, this advantage becomes even larger (up to about 30%) when the data center is located farthest from the center of the network topology.

This letter presents the architecture of multi-gigabit parallel demodulator suitable for demodulating high order QAM modulated signal and easy to implement on FPGA platform. The parallel architecture is based on frequency domain implementation of matched filter and timing phase correction. Parallel FIFO based delete-keep algorithm is proposed for timing synchronization, while a kind of reduced constellation phase-frequency detector based parallel decision feedback PLL is designed for carrier synchronization. A fully pipelined parallel adaptive blind equalization algorithm is also proposed. Their parallel implementation structures suitable for FPGA platform are investigated. Besides, in the demonstration of 2 Gbps demodulator for 16QAM modulation, the architecture is implemented and validated on a Xilinx V6 FPGA platform with performance loss less than 2 dB.